@jordancurve hm, that still looks slow

this is extremely weird, ngn/k for me is extremely slow, 900+ms for just 1+1

seems like there's a constant 800ms or so

@rak1507 it's instant for me. what compiler do you use and what version?

oh, not locally, the web editor

that too is almost instant for me

strange

at least the repl is instant

the timings it reports in the upper left corner are a bit strange though - randomly ranging from 2ms to 200ms

the timings are definitely right because I can feel the latency

@ngn s/upper left/upper right/

@rak1507 what browser do you use?

firefox

same here

ok, that's really weird, I just spammed some stuff a few times and then it dropped from ~500ms to ~3ms, so seems to work now

nope, back to slow speed...

firefox just crashed lmao I think something might just be wrong with my laptop

@jordancurve does k9 optimize *&~x? maybe something involving x?0 would be faster?

&n#~!p is another candidate for easy optimization

&n#~!p <-> p*!1+_(n-1)%p

@rak1507 i've had that happen to me because of plugin autoupdates

who knows what they're doing.. might be mining crypto coins or something

@ktye @dzaima fyi, i made some experiments with grading algorithms on modern hardware: c stdlib is slow (300+ ms, including its mergesort impl), hand-written pure mergesort is fast (200ms, the half-copying variety brings no measurable benefits), radix sort is crazy fast (100ms)

well, that was strange, but it all seems fine now

@ngn it's weird to me that the standard library would be slower than a hand written thing, is it not at all optimised?

@rak1507 idk. this is what i measure.

@rak1507 things in the C standard library are lacking in many aspects

tbf, stdlib has the disadvantage of having to call a non-inlined cmp() function, but even when i changed my own code to do the same, stdlib was still slower

(all timings i mentioned above are for grading 10^7 pseudo-random 64bit ints)

probably worth testing other sets of data - !1000000, reversed that, almost-sorted, smaller ranges, many/all duplicates, etc

the design of the c stdlib qsort() is really a pain in the ass, since there's no way to inject context into the comparator function except globals

if you're doing anything other than sorting an array of values the api is simply inadequate

@JohnE see qsort_r()

that's a gnu extension, though

ngn.bitbucket.io/k/#eJxLs6pWUlDS19DRUq/… can this be golfed?

should be fairly self explanatory